Nobuyoshi Nakada
Nobuyoshi Nakada
> Seems like this information was lost in the issue. As there were many relevant or similar issues, not all were taken care of.
On PPC64, doesn't `mftb` read 64-bit value? https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/tree/arch/powerpc/include/asm/vdso/timebase.h?id=0faa22f09caadc11af2aa7570870ebd2ac5b8170
> @nobu Yes, you could also use `mftb` + `mftbu`, but the code I used comes from the original code which that section of gc.c is based upon. Maybe the...
Maybe you could simplify more by https://github.com/ruby/ruby/pull/6092?
Aliasing to the same name is intentionally adding the count, to suppress the later method redefinition warning.
As Ripper is extensible by user classes inheriting `Ripper`, and can't know the dispatched results, I doubt that the way like "parser" gem, add `shadowargs` to params, is possible.
Or, `class Integer; def ceildiv(other) = -div(-other); end` in numeric.rb. This may be faster due to method caches.
Would you merge or rebase to fix the OpenSSL failures?
This seems fine to me, but we have no RISC-V and ARM32 FreeBSD CI machines.
You can add a benchmark script under `benchmark/`, like 5e897227ff3d37a36be96bb2c082370d437058ea, instead of writing in a commit log, and `-omarkdown` options makes a markdown table. # Iteration per second (i/s) |...